# Recording & Ingest

> Skip hand-authoring — record a narrated demo and let decodx ingest scaffold the project.

# Recording & Ingest

You don't have to author every scene by hand. Record yourself demoing with your voice on the track,
then let `decodx ingest` turn that recording into a ready-to-render project.

## Record

Use any screen recorder that captures your voice — QuickTime, OBS, or `Cmd+Shift+5`. Narrate as you
go; that narration becomes each scene's script. For clean results, pause briefly between steps and
record at your target resolution.

## Ingest

```bash
decodx ingest recording.mp4 --project my-video
```

Ingest transcribes the narration (with word timestamps), splits the recording into scenes at long
pauses, cuts one silence-trimmed clip per scene, and emits the whole YAML file set — each segment's
transcript becomes that scene's `narration`. The recording's own audio never reaches the output; the
normal pipeline re-voices it with your chosen voice.

Useful flags:

- `--slides` — scaffold editable intro/outro title slides.
- `--clean` — tighten each scene's narration to its clip budget with the cleanup LLM.
- `--zoom` — detect on-screen activity and emit editable `zoom` effects.
- `--update` — re-ingest while preserving scenes you've hand-edited.
- multi-take: `decodx ingest a.mov b.mov …` composes several recordings into one project.

Ingest generates authoring files and stops — it never runs the pipeline. Review and tighten the
narration, then `decodx run` renders the video.

Next: the [YAML Reference](../yaml/overview) to understand (and hand-tune) what ingest produced.